Abstract: For applications including dependent tasks with timing constraint in the case of considering transition overhead and inter-core communication overhead on variable voltage multi-core processors, an overhead-conscious synthesis method of energy optimization was proposed for real-time multi core embedded systems. The method effectively combined dynamic power management with adaptive body bias and dynamic voltage scaling based on independent tasks produced by the RDAG algorithm. Simulation experiments were conducted with several random task graphs and task graphs representing real applications on 2，3 cores, respectively. The experimental results show that the proposed method gets advantage over the original method.

Key words: Energy optimisation, Dynamic power consumption, Leakage power consumption, Multi-core